Evaluation of the sieve net as a selectivity-improving device in the Belgian brown shrimp (Crangon crangon) fishery

Fisheries Research 69 (2004) 35–48

Evaluation of the sieve net as a selectivity-improving devicein the Belgian brown shrimp (Crangon crangon) fishery

Hans Poleta,∗, Johan Coenjaertsb,1, Reinhart Verschoorec,2a CLO-Sea Fisheries Department, Ankerstraat 1, B-8400 Oostende, Belgium

b Centro de Estudios Avanzados de Blanes, c/Accés a la Cala St. Francesc, 14, 17300 Blanes, Girona, Spainc Department of Agricultural Engineering, Faculteit Landbouwkundige en Toegepaste Biologische Wetenschappen,

Ghent University, Coupure Links 653, 9000 Ghent, Belgium

Received 26 September 2003; received in revised form 14 January 2004; accepted 2 April 2004

Abstract

The fishery for brown shrimp (Crangon crangon) in the North Sea is carried out by more than 600 vessels, with totalannual brown shrimp landings of around 20,000 t. Due to the small mesh size used, the catches also contain large amountsof unwanted by-catch. To find ways of reducing this by-catch, experiments were carried out during a one-year period witha sieve net rigged in a brown shrimp beam trawl. It was tested for its selective properties and an evaluation was made of itsoperational characteristics.

The loss of commercial shrimp catch when using a sieve net was 15% or less in favourable conditions. Certain componentsof the catch can, however, lead to a distortion of the outlet with a reduction of the commercial catch of over 30%. It is likelythat technical alterations to the outlet can prevent this. The sieve net showed very poor selective properties for commercialfish species with a length below 10 cm. Above 10 cm the selection improved with increasing length. Especially for Age 1 andolder fish, this selective device serves its purpose. A significant reduction of non-commercial fish and invertebrates was alsoobserved.© 2004 Elsevier B.V. All rights reserved.

Keywords:Discards; Sieve net; Brown shrimp;Crangon; Beam trawl; North Sea

1. Introduction

The discard problem in the brown shrimp fisheryhas been recognised for a long time (e.g.Gilson,1935). Several early attempts were made to reducethe by-catch, but only by increasing the mesh size

∗ Corresponding author. Tel.:+32-59-265504;fax: +32-59-330629.E-mail addresses:[email protected] (H. Polet),[email protected] (J. Coenjaerts), [email protected](R. Verschoore).

1 Tel.: +34-972-336101; fax:+34-972-337806.2 Tel.: +32-9-264-61-29; fax:+32-9-264-62-35.

(Roelofs, 1950; Gilis, 1951; Mistakidis, 1958). Thisresulted in a reduction of undersized shrimps in thecatch, often accompanied with a reduction of thecommercial catch.

In 1965, the first results were reported on trials witha new selective device for brown shrimp trawls to re-duce the by-catches (Kurc et al., 1965). The device wasbased on the separator panel principle. Contrary to theabove-mentioned experiments, the aim was not only toreduce by-catches of undersized shrimps, but mainlyto reduce discarding of undersized fish. These trialswere soon followed by Dutch experiments (Boddeke,1965) with the same device and later on with an early

0165-7836/$ – see front matter © 2004 Elsevier B.V. All rights reserved.doi:10.1016/j.fishres.2004.04.007

36 H. Polet et al. / Fisheries Research 69 (2004) 35–48

version of the sieve net used nowadays. The reportconcluded that the catch of juvenile fish and smallshrimps was reduced, that the quality of the catch im-proved and that the workload on board was reduced.In the Netherlands, part of the shrimp fleet adoptedthe sieve net on a voluntary basis to improve selectiv-ity (Besançon, 1973). The report mentioned, as one ofthe disadvantages of the system, the loss of commer-cial shrimps when hydroid polyps were present in thewater and stated that the selective trawl should onlybe used in the absence of material that can clog thesieve net meshes. Comparable experiments were alsocarried out in Belgium (Van Middelem and Cleeren,1967) but in some cases a significant commercial catchloss occurred. Also in the US similar experimentswere carried out with promising results (Anon., 1968).Based on the fairly successful developments of sep-arator panels in shrimp trawls and the perception ofthe seriousness of the discard problem in shrimp fish-eries, FAO organised an expert consultation workshopon the issue (Anon., 1973). In France there were plansto enforce the sieve net (Brabant, 1973). The report,however, described the poor selectivity of the devicefor the smallest flatfish.

In the same period, experiments were carried outwith separator panels in the northernPandalusshrimpfishery, in Norway (Rasmussen, 1973) and Iceland(Thorsteinsson, 1973). Although the fishery had somedifferent characteristics, the issue was identical, sepa-rating fish from shrimps during the capture process. Asimilar device was also tested successfully as a turtleexcluder device in a shrimp trawl (Kendall, 1990). Inseveral North Sea areas, fishermen started using thesieve net on a voluntary basis. An example is Germanywhere the device is used dependent on area and sea-son (Mohr and Rauck, 1979). These authors clearlystated that application in the fishery was not based onprotecting young fish but on the practicalities for thefishermen, saving time and labour while sorting andcontinuing to fish when large quantities of jellyfishoccurred in the catches. By the start of the presentstudy, Denmark was the only country where the sievenet was legally enforced in the brown shrimp fishery(Van Marlen et al., 1997).

Based on the studies available, the sieve net seemsto be effective in releasing part of the discards, fish aswell as invertebrates. It is not made of rigid materialand therefore it is more acceptable to fishermen than

a rigid sorting grid. This, together with the observa-tion that fishermen tend to use the device voluntarilyin certain circumstances, makes the sieve net an ob-vious choice for further study. The sieve net wasstudied in the frame of the DISCRAN project, a co-operation of Bundesforschungsanstalt für Fischerei(Germany), the CLO-Sea Fisheries Department (Bel-gium), Rijksinstituut voor Visserij Onderzoek (theNetherlands—co-ordinator) and the University ofNewcastle (UK). This report focuses on the Belgianresults.

2. Materials and methods

2.1. Fishing area, vessels and trawls

The trials were carried out aboard the research ves-sel (RV) “Belgica” during two cruises in February–March and November 2000 and on the commercialshrimp trawler O.700, “Bisiti”, during 14 cruises be-tween April 2000 and January 2001. A total of 93hauls was carried out with the sieve net. The purposeof these trips was to evaluate an existing sieve netdesign in the Belgian shrimp fishery.

The A.962 “Belgica” is a governmental oceano-graphic research vessel, fishing from the stern. Eachtrip, a fishery skipper was hired to select fish tracksand guide the experiments to have practices as closeas possible to commercial conditions. It has to benoted, however, that due to safety regulations and re-stricted access of the research vessel to shallow fishinggrounds, the traditional shrimp fishing grounds couldnot be visited and the experiments were mainly carriedout in the area “Vlakte van de Raan” (flat, hard sandygrounds) giving a rather low number of species in thecatch.

The commercial vessel was specifically fitted forcatching shrimps (rotating shrimp riddle, cooker, etc.)and had a three-man crew. In contrast to the researchvessel, the commercial boat towed two beam trawlssimultaneously, one at each side of the vessel. For thesea trials, the choice of fishing grounds and fishingpractices was left to the skipper to ensure commer-cial conditions. The traditional fishing grounds on theFlemish Banks off the Belgian coast were fished (ICESsub-division IVc). In the area fished, a wide varietyof sediments is found (Bastin, 1974). Van Lancker

H. Polet et al. / Fisheries Research 69 (2004) 35–48 37

(1999)demonstrated that this is an area with typicallong stretched sand banks having irregular relief withvarying slopes and continuously dredged navigationchannels. A wide range of sand dunes, from small tolarge, occurs in this area. This large variation in sed-iments and topography affects the behaviour of thefishing gear.

The towing speed was between 2 and 3 knots onthe research as well as on the commercial vessel. Thevariability in towing speed was somewhat higher onthe commercial vessel due to the influence of tide butthe differences were quite small. The warp length wasthree times the water depth.

The gear studied (Fig. 1) was a commercial shrimpbeam trawl with a beam length of 8 m and a verticalnet opening of 0.5 m. The lengths of the headline andthe groundrope were 7.8 and 9.8 m respectively. Thebobbin rope consisted of rubber bobbins with a diam-eter of 21 cm rigged on steel axles with a diameter of20 mm. The net was made of knotted polyamide net-ting with nominal mesh sizes of 28 mm in the frontpart decreasing to 22 mm in the aft part. The cod-endwas made of knotted polyamide netting with a nomi-nal mesh size of 22 mm and protected by a polyamidelifting bag with a nominal mesh size of 80 mm.

Sieve nets are rarely used in the Belgian shrimpfishery (Van Marlen et al., 1997). For the design of thesieve net to be used in the Belgian trials, the net plansused by the Dutch (for a 9 m beam) and the UK (for a6 m beam) project partners were used and scaled to thesize of an 8 m beam. Both sieve net designs were beingused in the commercial fishery in the Netherlands andthe UK at the time of the project. After scaling, bothsieve nets were almost identical. The position in thenet was copied from the other partners. The net planand the positioning are given inFig. 1.

The sieve net had a nominal mesh size of 70 mm,was 116 meshes wide at the front, 16 meshes wide atthe rear and 60 meshes deep. The outlet of the sievenet was positioned in the belly of the net, close to thecod-end.

As the sieve net was already in use in the com-mercial fishery, no further design and optimisationwork was done. This design was used throughout thetrials to test its performance in different conditions.The trials were grouped according to the season andwhether they were carried out on a commercial or aresearch vessel.

2.2. The cod-end and the outlet covers

The cod-end used was a standard commercialcod-end. Aboard RV Belgica, only one gear couldbe towed. Therefore, the cod-end of the experimentalnet was blinded with 11 mm mesh netting and theoutlet of the sieve net was covered with an 11 mmoutlet cover to retain the escapees from the outlet. Forcatch comparison aboard the commercial vessels, theexperimental net was towed from the starboard sideof the vessel while the standard net was towed fromthe port side. At regular time intervals the skipperswere asked to check during the commercial fisherywhether the two sides of the vessel had the samecatching efficiency, to avoid bias. This informationwas not analysed statistically but was used as a rou-tine quality check for the performance of the fishinggears. The catching efficiency of both sides was verysimilar. Commercial mesh sizes were used. Unlikethe shrimp fishery elsewhere in the North Sea, theBelgian shrimpers catch marketable sized fish and areallowed to land these fish. Financially the Belgianshrimp fleet depends strongly on the income from fishlandings. The use of a sieve net that directs the largerelements of the catch towards an outlet would lead toa substantial loss of commercial catch. Therefore, thetrials aboard commercial vessels were always carriedout with an outlet cover. This cover had a nominalmesh size of 80 mm, which is the legal minimummesh size for the sole fishery. The cover was attachedaround the outlet at a sufficient distance from theoutlet to avoid distortion, although problems werenot expected since catch volumes in the cover wouldusually be quite low because of the large mesh size.

The average cod-end mesh openings were 21.5and 21.8 mm with standard deviations of 0.62 and0.90 mm. The average mesh openings of the coverand blinder were 10.9 and 10.6 mm with standarddeviations of 0.63 and 0.78 mm, respectively. Theaverage mesh opening of the large mesh outlet coverwas 77.9 mm with a standard deviation of 2.30 mm.

2.3. Data collection

The species investigated were brown shrimp (Cran-gon crangon), plaice (Pleuronectes platessa), sole(Solea solea), dab (Limanda limanda), flounder(Platichtys flesus), whiting (Merlangius merlangus),


Fig. 1. The experimental net rigged with a sieve net and the net diagram.

bib and poor cod (Trisopterus luscusandTrisopterusminutus; not discriminated) and cod (Gadus morhua).Only those species present in sufficient numbers inthe catch were included in the data analysis. Besidesthese commercial species, non-commercial fish andinvertebrates were also studied.

On board, the fishermen processed the catch througha rotating shrimp riddle. After sorting, the catch wassplit into three fractions:

• the main by-catch containing the larger fishes, crabs,starfish, debris, etc.,

• the commercial shrimps with 0- and 1-group flatfishtogether with occasional juvenile roundfish and

• the non-commercial shrimps with the smallest flat-fish.

The volume of each catch fraction was recorded.This allowed a preliminary assessment of the func-


tioning of the sieve net. Further analysis was done onnumbers of individuals by species.

The fish in the catch were sorted out and measuredimmediately after the haul. The fish were measuredfor total length (TL) to the centimetre below, fromthe tip of the nose to the tip of the stretched tail fin.This method was preferred over standard length (i.e.from the tip of the nose to the fork of the tail fin),to be in accordance with the method used to measurefish for their compliance with minimum landing sizeregulations.

For shrimp, samples were taken for later analysis inthe laboratory. There, the total length of the shrimpswas measured to the millimetre below from the tip ofthe scaphocerite to the distal margin of the fans on thestretched uropods. An image analysis method (devel-oped by Bfafi, Hamburg, D) was used to measure theshrimps. Following the results of a theoretical studyon the effect of sample sizes on the estimation ofselection parameters for shrimp trawl cod-ends (Poletand Redant, 1999), it was decided to measure at least250 shrimps per catch fraction.

To estimate the amount of non-commercial fish andinvertebrates from each haul, a sample was taken. Inthe laboratory the animals were weighed and counted.

2.4. Data analysis

The proportion of shrimps and fish escaping throughthe outlet was calculated differently for the researchand the commercial vessel:

RV: pe(l) = catch in outlet cover/(catch in outletcover+ catch in main cod-end withblinder)

CV: pe(l) = (catch in control cod-end− catchin experimental cod-end)/catch incontrol cod-end

with pe(l) the percentage escaping at length. If pos-sible, a sieve net selection ogive was fitted to thefish length selectivity data. The method used for thispurpose follows the standards laid out in the “ICESSelectivity Manual” (Wileman et al., 1996). The linkfunction used for individual hauls was the logisticcurve and gives the probability pe(l) that a fish oflength l escapes, given that it entered the sieve net:

pe(l) = exp(a + bl)

1 + exp(a + bl)

wherea and b, the two parameters to be estimated,represent the intercept and the slope, after a logittransformation. These parameters were estimatedwith the maximum likelihood method using the CCsoftware (Constat, Denmark). Estimation ofa and band the variance matrix is described inFryer (1991).L25, L50 and L75 are the body lengths at which 25,50 and 75% of the brown shrimps escape. SR is theselection range, equal to the difference between L75and L25 and gives an idea of the slope of the curve.

L50 = −a

b, SR= 2 loge (3)

b,

L25 = L50 − SR

2, L75 = L50 + SR

2

In the case of the research vessel trials, where a coverwas used, the analysis followed the procedure forcover data (Wileman et al., 1996). For the commercialvessel, the data were obtained from paired gear ex-periments. In this case, the methodology applied wasthe “share each lengths catch total” (SELECT) sta-tistical model developed byMillar and Walsh (1990,1992). For these paired gear data, in addition toaand b an estimate ofp, the relative fishing powersor efficiencies of the two nets, is obtained using theSELECT model. The paired gear method is based onthe assumption that the control cod-end of the stan-dard gear is non-selective for fish, i.e. all fish in theselection span of the test cod-end are retained. Thiswas the case since the shrimp beam trawl is almostnon-selective for fish (Polet, 2000).

The selectivity parameters of single hauls werecombined to calculate a mean curve, taking into ac-count between haul variation, by the variance compo-nent analysis method laid out byFryer (1991), usingthe CC software.

For the comparison of catch differences, at-test forindependent samples was used and backed up with thenon-parametric Mann–WhitneyU-test if samples weresmall and normality of the data was not convincing.

3. Results

3.1. Narrative of the sea trials

The first trials with a sieve net were conductedaboard RV “Belgica” for a preliminary assessmentof its selective properties, at the end of February and


Table 1Log of the experimental hauls with the sieve net

Code Vessel Period Number ofhauls

Rigging Action

RV-Trials RV Belgica February–March andNovember 2000

21 Standard sieve net Grid selectivity—covermethod

CV-Spring O.700 April 2000 15 Standard sieve net Grid selectivity—catchcomparison

CV-Summer O.700 July and September 2000 20 Standard sieve net Grid selectivity—catchcomparison

CV-Autumn–Winter O.700 November and December2000, January 2001

37 Standard sieve net Grid selectivity—catchcomparison

the start of March 2000. The first series of trials witha sieve net aboard a commercial vessel (CV) wereconducted in April, referred to as “CV-Spring” trials.The next series of sea trips was carried out in Julyand September, referred to as “CV-Summer”. The“CV-Autumn–Winter” trials were executed betweenNovember and January. A second trial was performedon RV “Belgica” in November to check whether thegood results obtained on the research vessel in theoff-season could be replicated in a season when largecatches prevail. As these results were comparable,the experiments were grouped and referred to asRV-Trials. An overview is given inTable 1.

The average haul duration was 1 h and 33 min witha minimum and maximum of 15 min and 2 h and50 min, respectively. In each haul, sufficient numbersof shrimps were caught for data analysis. The num-bers of fish in the catch varied strongly, dependingon the fish species, time of day, season and fishingground. For many hauls, though, the numbers caughtwere sufficient to calculate selectivity ogives at haullevel. This was the case for dab, plaice, bib and whit-ing (except for bib in CV-Spring). For each seriesof experiments and each fish species, the selectivityparameters and corresponding confidence limits werethen calculated for the combined hauls. Cod was onlycaught during the RV-Trials. Due to the low numbersin the catches, the selectivity ogive was calculatedon the pooled hauls. Sole did not show the normalselection pattern as observed for the other species andthe typical ogive could not be fitted.

3.2. Sieve net selectivity

On the commercial vessel, the experimental netcaught less compared to the standard net for each of

the three “unsorted catch fractions”, i.e. on averageover all hauls 40% (29–51%) less main by-catch,44% (32–55%) less discard shrimp fraction and 23%(15–30%) less commercial shrimp fraction. On theV the results were much better, with less than 8%(7–8%) loss of commercial shrimps, 17% (15–20%)loss of discard shrimps and a reduction of the mainby-catch fraction with 28% (24–33%).

3.2.1. Crangon crangonBased on the “numbers of shrimps” in the catches,

the RV-Trials gave very promising results, with only7% loss of the commercial shrimps and a catch re-duction for the discard shrimps of 14% (Table 2).These catch reductions were considered to be low,but still highly significant (P < 0.01). The sea tripsin CV-Spring more or less confirmed these resultswith again a rather low loss of the commercial shrimpfraction (13%) and a 34% reduction of the discardshrimps. The difference with the RV-Trials was,however, highly significant. The losses of shrimpsin CV-Summer were somewhat higher compared toCV-Spring, i.e. 15 and 35% for the commercial anddiscard shrimps respectively. The difference with the

Table 2The average percentages shrimps lost due to the sieve net selectionfor the fraction smaller than 45 mm, the fraction larger than 45 mmand for all length classes (95% confidence limits within brackets)

Data set %Lost(<45 mm TL)

%Lost(≥45 mm TL)

%Lost(all shrimps)

RV-Trials 14 (10–18) 7 (6–8) 10 (8–13)CV-Spring 34 (28–41) 13 (8–19) 20 (15–25)CV-Summer 35 (24–47) 15 (12–18) 26 (19–33)CV-Autumn–

Winter58 (52–65) 31 (28–35) 44 (39–48)


Fig. 2. Percentage ofC. crangonlost with the 95% confidence limits indicated as error bars.

RV-Trials was highly significant but not significantcompared to CV-Spring. The results for the autumnand winter trials were consistently different fromthe other commercial trials and the RV experiments,with very high losses of shrimps. An average loss ofcommercial shrimp catch of over 30% was observed.Also the catch reduction of the discard shrimps wentup to 58%. The differences with all other trials werehighly significant. For each of the differences, theMann–WhitneyU-test confirmed thet-test results.

The percentages of lost shrimps by length classis given in Fig. 2. For the RV-Trials, escapes werelow over all length classes. A length effect was quiteclear but not pronounced, with somewhat more smallshrimps escaping. For the CV-Spring and CV-Summertrials, losses of shrimps were more evident. For

lengths above 50 mm, a length effect was not clearand catch reductions were roughly 10%. Below thislength, though, losses increased with decreasing sizeof the animals and even exceeded 50% for CV-Spring.For the Autumn–Winter trials, a length effect wasobserved over the whole length range. For the largestanimals, losses were close to zero. Contrary to theCV-Spring and CV-Summer trials, the catch reduc-tion decreased steadily from 60% for the smallestshrimps to almost zero for the largest shrimps.

3.2.2. Commercial fish speciesThe selective properties of the sieve net for fish

in the different trials are given inTable 3 and pre-sented graphically inFig. 3(RV-Trials) andFig. 4(allRV-Trials combined). Note that the figures give the


Tabl

e3

Sie

vese

lect

ion

para

met

ers

for

fish

(95%

confi

denc

elim

itsw

ithin

brac

kets

)

Spe

cies

RV

-Tria

lsC

V-S

prin

gC

V-S

umm

erC

V-A

utum

n–W

inte

r

Sie

veL5

0(m

m)

Sie

veS

R(m

m)

Sie

veL5

0(m

m)

Sie

veS

R(m

m)

Sie

veL5

0(m

m)

Sie

veS

R(m

m)

Sie

veL5

0(m

m)

Sie

veS

R(m

m)

Dab

11.3

(10.

7–12

.0)

8.6

(6.2

–10.

9)9.

0(8

.1–9

.7)

6.4

(4.5

–8.3

)9.

7(7

.7–1

2.9)

6.8

(4.6

–9.0

)11

.6(9

.9–1

3.2)

7.0

(6.0

–8.1

)P

laic

e10

.0(9

.4–1

0.5)

5.9

(4.6

–7.3

)9.

0(7

.3–1

0.8)

8.3

(5.6

–10.

9)10

.2(7

.5–1

5.9)

4.9

(0.2

–9.7

)11

.1(9

.9–1

2.2)

5.3

(3.9

–6.8

)B

ib14

.9(1

3.0–

16.6

)8.

6(6

.2–1

0.9)

(–)

(–)

11.7

(7.8

–13.

6)11

.5(5

.9–1

7.1)

14.6

(12.

8–15

.9)

8.4

(6.3

–10.

4)W

hitin

g22

.7(2

1.5–

24.2

)13

.0(1

1.0–

15.0

)18

.8(1

7.7–

20.1

)12

.2(9

.3–1

5.2)

18.7

(15.

7–22

.0)

8.3

(6.0

–10.

6)20

.3(1

8.6–

22.0

)13

.0(1

0.9–

15.0

)C

od19

.2(–

)3.

9(–

)(–

)(–

)(–

)(–

)(–

)(–

)

percentages fish escaping which is the opposite of per-centages retained usually given by selectivity curves.

The selective properties of the sieve net for eachof the different fish species were comparable betweenthe research and the commercial vessels and over theseasons (Table 3). Although the catch reductions forshrimps were variable over the different trials, the se-lectivity of the sieve net for fish was more constant.

The sieve net was very selective for marketablefish. The marketable dab, plaice, sole, bib and codcatch was almost entirely sorted out (Figs. 3 and 4).Only marketable whiting succeeded in penetrating thesieve net meshes to some extent with retention in thecod-end of less than one quarter of the fish. The 80 mmoutlet cover retained most of the marketable fish afterthey escaped through the outlet. The loss of mar-ketable fish due to selection of the 80 mm outlet coverwas low during the Summer and Autumn–Wintertrials with 9 and 19% loss respectively. The lossesduring the spring trials were higher at 35%.

The sieve net showed a poor selection for all com-mercial fish species with lengths below 10 cm (i.e.dab, plaice, sole, bib, whiting and cod) (Figs. 3 and 4).Usually less than one-quarter of these animals man-aged to escape through the outlet. Above 10 cm, se-lection improved and a higher proportion of the fishcould escape. For the flatfish species dab and plaice,the selectivity was better compared to the roundfishspecies bib, cod and especially whiting that showed apoor selectivity.

The sieve net had very similar selective propertiesfor dab and plaice (Fig. 4). Below 10 cm length thecatch reduction was very low. The selection rose verysteeply between 10 and 15 cm to reach 100% catchreduction around 20 cm. For sole, the selection pat-tern deviated from the other flatfish species. For thesmallest animals, the sieve net sorted out a rather highproportion. Around 10 cm, the number of escapeesdropped to a minimum and increased again for thelarger length classes. One hundred percentage catchreduction was only reached at about 25 cm length.

The smallest bib (about 10 cm) could easily passthrough the sieve net meshes into the cod-end. Abovethat length selection improved and reached 100%around 25 cm length. For whiting a much shallowerselection pattern was observed. As for bib, almostall fish of 10 cm length passed through the sieve netmeshes but also many fish above 20 cm ended up in


Fig. 3. Sieve net selection ogives (black lines for single hauls, light blue line for combined hauls) and length frequency distributions (deepblue line) for dab, plaice, sole, bib, whiting and cod for the RV-trials. (For interpretation of the references to colour in this figure legend,the reader is referred to the web version of this article.)

the cod-end. The 100% retention was only reachedabove 30 cm. For cod, the data were scarce and theresults should be interpreted with caution. The se-lection ogive was quite steep compared to the otherroundfish species but L50 was comparable to the L50of whiting.

3.2.3. Non-commercial fish species and invertebratesThe average reduction of the unsorted main

by-catch fraction (mainly containing invertebrates anddebris), due to sieve net selection, on the RV-trialswas 28% (24–33%). The reduction for the CV-Spring,CV-Summer and CV-Autumn–Winter trials was 49%(44–55%), 29% (16–41%) and 41% (34–48) respec-

tively. The catch reductions of non-commercial fishand invertebrates at species level are given inFig. 5.For clarity, the results for all commercial hauls havebeen pooled and are presented together with theresults for the RV-Trials.

4. Discussion

Based on the amounts of lost shrimps, the experi-ments with the sieve net can be sub-divided into threegroups: (a) all RV-Trials—very low shrimp loss, (b)CV-Spring and CV-Summer trials—low shrimp loss,(c) CV-Autumn–Winter trials—high shrimp loss. The


Fig. 4. Sieve net selection ogives for dab, plaice, whiting and cod and retention points for sole—all hauls combined over all sieve netexperiments on the commercial vessel.

high shrimp losses for the latter came as a surprisesince the previous commercial vessel experiments andthe RV-Trials gave good results, which were quitesimilar to the results obtained by the other projectpartners (Van Marlen et al., 2001). For a number ofhauls, the poor operation of the sieve could be at-tributed to clogging with seaweed and hydroids butthis was not always the case. Throughout the trials, the

Fig. 5. The percentage reduction of the different non-commercial fish and invertebrate species in the catch of the experimental net (allCV-Trials combined and RV-Trials).

same net and sieve net were used, without alterations.In addition, the commercial vessel and its skipper andcrew were the same throughout the trials. The reasonfor this sudden change in commercial shrimp loss wasnot obvious. For a number of hauls with high com-mercial shrimp losses, the skipper aboard the fishingvessel noticed gilled fish in the sieve net meshesaround the outlet after hauling the net. Quantification


Fig. 6. The relation between the loss of shrimps (commercial and non-commercial) and the numbers of large (>30 cm) dab and whitingin the CV catch.

and measurement of these fish was not possible sinceafter bringing the net on board, these fish had usu-ally already dropped out of the meshes. Based on theexperience of the skipper, however, it was estimatedthat these fish were large (>30 cm) dab and whiting.The presence of these fish around the outlet wouldquite likely distort the outlet of the sieve net. Pos-sibly a higher water flow was then directed throughthe outlet, taking part of the catch with it. In orderto check this hypothesis, the correlation between thecatch reduction of shrimps and the numbers of largedab and whiting was calculated. The relation betweenboth variables is presented graphically inFig. 6.

From this graph, it is obvious that no clear lin-ear relation exists between the variables. What it doesdemonstrate, however, is that if there are no large daband whiting in the catch, there is a low probability forloss of shrimps (<20%). Higher numbers of these fishlead to a higher probability for loss of shrimps (>30%).

The early experiments done with sieve nets(Boddeke, 1965; Kurc et al., 1965), in the Netherlandsand France, did not indicate reductions of commercialcatch. The Belgian trials, however (Van Middelemand Cleeren, 1967), did pose problems with catchlosses of commercially sized shrimps for three of thesix sieve net designs tested. The authors could notexplain why but it was suggested that the problemsmight have been caused by the design of the net,although the differences in design were quite small.The catch composition was not studied in detail. Pos-sibly, the causes of the catch losses were the same as

in the present study but due to a lack of historic data,this cannot be checked. It is striking that both in the1960s and in the 1990s, the application of the sievenet was problematic in the southern North Sea, whileit worked well on the other shrimp grounds.

Revill et al. (2000)tested four designs of sieve netand found a length effect for brown shrimps. For eachdesign, the losses of undersized shrimps were highercompared to commercial shrimps. For all experimentswith a sieve net during the present study, a compa-rable length effect was observed for the catch reduc-tions of shrimps, with higher losses for the smalleranimals (Fig. 2). This effect may partly be explainedby slightly better cod-end selectivity in the cod-endof the experimental trawl. This may also be the rea-son why the retention points for sole in the lowerlength classes were unexpectedly high. The effect forshrimp was small for the RV-Trials, medium for theCV-Spring and CV-Summer trials and large for theCV-Autumn–Winter trials. A Canadian study withshrimp trawls (Anon., 1996) demonstrated that thewaterflow in the net can be considerably altered byinserting a selective device in the trawl. By insertinga sieve net in the shrimp trawl and cutting an outlet inthe belly of the net, part of the water flow is deviatedtowards the outlet while fishing (though minor dueto the high mesh size of the sieve net). This watercurrent takes part of the catch with it through the out-let. It is plausible that larger, heavier shrimps are noteasily deviated from their track by this flow. Smallershrimps, on the other hand, possibly drift more easily


with the waterflow. This could explain the length ef-fect observed and the difference in length effect in thedifferent trials. Why this effect was not observed forfish is unclear. Shrimps are rather passive in the trawland are taken by the water flow (Polet, 2000). Fishon the other hand are active swimmers and due to theabsence of herding in beam trawls these animals arenot exhausted by the time they reach the outlet (withinseconds). This difference in behaviour may be thereason why a length effect was not observed for fish.

During the RV-Trials, the sieve net worked properlyand no clogging was observed. In this case the waterflow through the outlet could be expected to be mini-mal as were the shrimp losses. The presence of a smallmesh outlet cover also may have further reduced thiswater flow. During the CV-Spring and CV-Summertrials, the catch losses (and the length effect) werehigher, probably because of some degree of cloggingcausing a somewhat higher water flow through the out-let. During the CV-Autumn–Winter trials, catch losses(and the length effect) were highest due to cloggingand gilled fish causing a higher water flow through theoutlet. Consequently, the length effect as observed dur-ing the trials strengthens the hypothesis that the mal-functioning of the sieve net was caused by clogging.

The small mesh outlet cover used during theRV-Trials may have masked the outlet giving a pos-sible underestimation of the selection process in thesieve net. The results for the RV-Trials should there-fore be interpreted with caution and catch reductionsmay have been underrated.

Most of the marketable fish were selected out bythe sieve net and were led through the outlet. Withoutoutlet cover, all these fish would escape and be lostto the fishermen. Therefore, it is considered essentialthat any future legislation will allow a large mesh out-let cover to be fitted over the outlet to retain the mar-ketable fish and reduce any possible loss of income tothe fishermen.

The sieve net has proved to be unable to save asignificant amount of fish below 10 cm, i.e. mainlyAge 0 fish. This was also found by the project part-ners (Van Marlen et al., 2001) and in early French(Brabant, 1973) and German experiments (Mohr andRauck, 1979) and stated as a major disadvantage ofthe sieve net. Above 10 cm, selection improves anda large proportion of the Age 1 and older fish escapethrough the outlet. A biological and economic mod-

elling exercise (Revill et al., 1999; Revill, 2000) hasshown that for the typical Belgian catch compositionin the shrimp fishery, a reduction in discarding ofAge 1 and older fish can benefit the fish stocks. Thepossible benefits of saving Age 0 fish were consid-ered negligible in this area. Therefore, despite the lowselectivity of the sieve net for Age 0 fish, this devicewould be valuable if adopted by the shrimp trawlerfleet in Belgium, and less so in the Wadden Sea areawhere Age 0 fish are predominantly present.

The reduction of the unsorted main by-catch frac-tion as a whole (mainly containing invertebrates anddebris) on the commercial vessel, by sieve net se-lection, was significant. It was highest in Spring,Autumn and Winter and close to 50%. In summer,this catch reduction was lower (29%), caused by theincidental capture in a number of hauls of very highnumbers of unusual small crabs (Liocarcinus hol-satus), small enough to pass through the sieve netmeshes into the cod-end. This catch decreased the av-erage catch reduction. Excluding these hauls gave anaverage catch reduction of the main by-catch fractionof 53% (45–61%), which was comparable with theother commercial vessel trials. At species level, thecatch reduction of invertebrates and non-commercialfish varied quite strongly, between 23 and 100%. Forthe majority of the species, though, this reductionwas higher in the commercial vessel trials comparedto the catch reductions observed for the same speciesduring the RV-Trials. This also was the case for thereduction of the unsorted main by-catch fraction. Pos-sibly, the absence of clogging during the RV-Trialsand masking of the outlet by the small mesh covermay have caused these lower discard reductions.

The losses of commercial shrimps with the sievenet, as observed in commercial conditions in springand summer, were significantly lower compared to thelosses obtained with a selective sorting grid in the Bel-gian brown shrimp fishery (Polet, 2003). The selectiveproperties for fish of grids and sieve nets are differ-ent. For the sieve net, selectivity was constant over theseasons and trials. With the grid, the results were veryvariable and probably influenced by clogging.

The sieve net has proved to be an easy to fit androbust device. Since it had already been used in certainareas in the commercial shrimp fishery, it was wellknown in the fisheries community. The fishermen hada rather neutral attitude towards this device. As any


alteration to the traditional shrimp beam trawl to in-crease selectivity would probably complicate the fish-ery and increase costs and labour, fishermen generallydislike selective devices. Several contacts with thefishermen, however, demonstrated that the sieve netcould be acceptable. It was feared, though, that clog-ging in certain seasons could be a problem leadingto loss of commercial catches and fishing time. Theexperiments showed that loss of commercial shrimpsis indeed inevitable with the sieve net. These catchreductions usually exceeded 10% and in cases ofclogging or presence of gilled fish could exceed 30%.It can, however, be expected that technical alterationsto the sieve net outlet (different netting material,smaller mesh size) could reduce the clogging problemsignificantly.

5. Conclusions

During a 1-year period, a sieve net was tested forits selective properties and an evaluation was made ofits operational characteristics. The loss of commer-cial shrimp catch when using a sieve net was 15%or less in favourable conditions. Certain componentsof the catch can, however, lead to a distortion of theoutlet with a reduction of the commercial catch ofover 30%. It is likely that technical alterations to theoutlet can prevent this. The sieve net showed verypoor selective properties for commercial fish specieswith a length below 10 cm. Above 10 cm the selectionimproves with increasing length. Especially for Age1 and older fish, this selective device serves its pur-pose. The sieve net was less susceptible to cloggingcompared to a grid and performed better in differentconditions. It was rarely subject to damage. Based onthe results of a biological and economic modellingexercise the application of the sieve net in the Belgianshrimp fishery can be considered as a valuable toolto reduce the impact of discarding in this fishery to aminimum. On top of this, the application of the sievenet also leads to a significant reduction in unwantedby-catch of invertebrates and non-commercial fish,which would reduce the impact of the shrimp fisheryon the marine environment in general. It is, however,important to acknowledge that losses of commercialshrimp in certain seasons and areas, lead to financiallosses for the shrimp fishermen.

Acknowledgements

The authors wish to thank the skipper and the crewof the vessel O.700 Bisiti for their co-operation. Thisstudy was partly funded by the European Commission(Biological Study—contract no. 98/012).

References

Anon., 1968. New shrimp sorting trawl looks good in tests. FishBusiness 3 (10).

Anon., 1996. Construction and testing of a 1:2 model 1168 TerraNova shrimp trawl cod-end with selectivity devices. FishingTechnology Unit Report No. 6/96. Memorial University ofNewfoundland.

Anon., 1973. Report of the expert consultation on selective shrimptrawls—Ijmuiden, The Netherlands, 12–14 June 1973. FAOFisheries Report No. 139.

Bastin, A., 1974. Regionale sedimentologie en morfologie van dezuidelijke Noordzee en van het Schelde Estuarium. UnpublishedPh.D. Thesis. Katholieke Universiteit Leuven, Belgium.

Besançon, 1973. Review on the development of the selectiveshrimp trawl in the Netherlands. In: Anon., 1973. Report of theexpert consultation on selective shrimp trawls—Ijmuiden, TheNetherlands, 12–14 June 1973. FAO Fisheries Report No. 139.

Boddeke, R., 1965. Een beter garnaalnet. Visserijnieuws 18 (1),2–8.

Brabant, J.C., 1973. Devismes’s selective trawl for brownshrimp (Crangon crangon). In: Anon., 1973. Report of theexpert consultation on selective shrimp trawls—Ijmuiden, TheNetherlands, 12–14 June 1973. FAO Fisheries Report No. 139.

Fryer, R.J., 1991. A model of between-haul variation in selectivity.ICES J. Mar. Sci. 48, 281–290.

Gilis, C., 1951. L’effet de tamissage des mailles du chalut àcrevettes. Annales Biologiques Belgium VIII, 173–174.

Gilson, G., 1935. Recherches sur la destruction du jeune poissonpar la peche crevettière sur les cotes de Belgique. Mémoire no.3, Annales de l’Institut d’ Etudes Maritimes, Belgium.

Kendall, D., 1990. Shrimp retention characteristics of the Morrisonsoft TED: a selective webbing exclusion panel inserted in ashrimp trawl net. Fish. Res. 9, 13–22.

Kurc, G., Faure, L., Laurent, T., 1965. Essaisd’un nouveau typede chalut équipé d’un dispositif de sélectivité pour la pechedes crevettes. Science et Peche, Bulletin d’information et dedocumentation de l’Institut Scientifique et Technique des PechesMaritimes, April 1965, No. 136.

Millar, R.B., Walsh, S.J., 1990. Analysis of trawl selectivity datawith an application to trouser trawls. ICES C.M. 1990/B:14.

Millar, R.B., Walsh, S.J., 1992. Analysis of trawl selectivity studieswith an application to trouser trawls. Fish. Res. 13, 205–220.

Mistakidis, M.N., 1958. Comparative fishing trials with shrimpnets. Fishery Investigations, series II, vol. XXII, No. 1. Ministryof Agriculture, Fisheries and Food, UK.

Mohr, H., Rauck, G., 1979. First results of German experimentswith a selective shrimp trawl. ICES C.M. 1979/B:7.


Polet, H., Redant, F., 1999. Effect of population structure, samplingstrategy and sample size on the estimates of selection parametersfor shrimp (Crangon crangon) trawls. Fish. Res. 40, 213–225.

Polet, H., 2000. Cod-end and whole trawl selectivity of a shrimpbeam trawl used in the North Sea. Fish. Res. 48, 167–183.

Polet, H., 2003. Evaluation of by-catch in the Belgian brownshrimp (Crangon crangonL.) fishery and of technical meansto reduce discarding. PhD Thesis. Ghent University, Belgium.

Rasmussen, B., 1973. Fishing experiments with selectiveshrimp trawl in Norway. In: Anon., 1973. Report of theexpert consultation on selective shrimp trawls—Ijmuiden, TheNetherlands, 12–14 June 1973. FAO Fisheries Report No. 139.

Revill, A., Pascoe, S., Radcliffe, C., Riemann, S., Redant, F., Polet,H., Damm, U., Neudecker, T., Kristensen, P., Jensen, D., 1999.The economic and biological consequences of discarding inthe EuropeanCrangonfisheries. Final Report to the EuropeanCommission, Contract No. 97/SE/025.

Revill, A., Riemann, S., Radcliffe, C., Dutton, K., Bower, S.,Jeffery, S., Frid, C., Taylor, K., 2000. The discarding of nonmarketable (undersized)Crangon crangonin the UK Crangoncrangonfishery. Final Report to the Chief Scientist Group ofthe Ministry Agriculture, Fisheries and Food (London, UK),MAFF CSG Project No. MF 0615.

Revill, A., 2000. Discarding in a European fishery and its effectupon the spawning stock biomasses, lost landings and theimplications for corrective selective measures. In: Meeting ofthe ICES Working Group on Fishing Technology and FishBehaviour, Haarlem, the Netherlands, 10–14 April 2000.

Roelofs, E.W., 1950. Mesh size and construction as a factor inreleasing small fish and shrimp from trawl nets. US Fish andWildlife Service, Fisheries Leaflet 382.

Thorsteinsson, 1973. Selective shrimp trawl experiments inIcelandic waters. In: Anon., 1973. Report of the expertconsultation on selective shrimp trawls—Ijmuiden, TheNetherlands, 12–14 June 1973. FAO Fisheries ReportNo. 139.

Van Lancker, V., 1999. Sediment and morphodynamics of asiliciclastic near coastal area, in relation to hydrodynamicaland meterological conditions: Belgian continental shelf. PhDThesis, Ghent University, Belgium.

Van Marlen, B., Redant, F., Polet, H., Radcliffe, C., Revill, A.,Kristensen, P.S., Hansen, K.E., Kuhlmann, H.J., Riemann, S.,Neudecker, T., Brabant, J.C., 1997. Research into theCrangonfisheries unerring effect. RIVO-DLO rapport C054/97

Van Marlen, B., de Haan, D.I., Revill, A.S., Dahm, E., Wienbeck,H., Purps, M., Coenjaerts, J., Polet, H., 2001. By-catchreduction devices in the EuropeanCrangon fisheries. ICESC.M. 2001/R:10.

Van Middelem, A., Cleeren, G., 1967. Een onderzoek vangarnaalnetten in Beneluxverband. Werkgroep “Techniek inde Zeevisserij”. Proefstation voor Zeevisserij, rapport nr. 24,1967.

Wileman, D.A., Ferro, R.S.T., Fonteyne, R., Millar, R.B. (Eds),1996. Manual of Methods of Measuring the Selectivity ofTowed Fishing Gears. ICES Cooperative Research Report. No.215.

Evaluation of the sieve net as a selectivity-improving device in the Belgian brown shrimp (Crangon crangon) fishery

Documents